4
$\begingroup$

How can I create a random group of order $4096$ with a center of size $2$ ?

The algorithm should be able to create every possible group with the given properties in principle. I think the list of groups with the required properties is far too large and probably not even known.

I tried semidirect products. But first, I can produce only a small part of the possible groups with that and second, either there are too many homomorphisms (and I do not know whether GAP can choose a random homomorphism) or the groups I get have a center with a size larger than $2$.

I did not find a command in GAP constructing a random group of a given order, if the list of all groups is not available.

Any suggestions ?

$\endgroup$
  • $\begingroup$ "Random group of order $\;2^{12}\;$"? If you specify that its center is of size two, how random do you want it to be? The number of groups of order $\;4,096\;$ seems to be at least in the hundreds of thousands of millions, and very probably in the trillions. "Randomly" can be a rather pretty long shot to produce a group as the one you want. $\endgroup$ – DonAntonio Mar 29 '16 at 8:24
  • $\begingroup$ Do you mean that groups of order $4096$ have a center of size $2$ very rarely, and therefore it would take very long to get a group with $|Z|=2$ by just choosing random groups of order $4096$ ? $\endgroup$ – Peter Mar 29 '16 at 8:28
  • $\begingroup$ No, I didn't say so. In fact, I've no idea what the size of the centers of those groups can possibly be (except that it cannot be of size $\;2^{11}\;$ as then $\;G/Z(G)\;$ would be cyclic non-trivial and this is impossilbe always). What I meant is that there are so many groups of order $\;4,096\;$ that taking randomly one and expecting it will have a center of size $\;2\;$ can be a very long shot unless something more is known, or conditioned, on such groups. $\endgroup$ – DonAntonio Mar 29 '16 at 8:35
  • $\begingroup$ I tried to get know more precisely the meaning of question. Let me say this, and know which direction you want to go. Consider two groups of order $2^6$ with center of order $2$ and take their central product. You may replace groups of order $2^6$ by one of order $2^5$ and other of order $2^7$,or one of order $2^4$ and other of order $2^8$, or.....(in short: take central products of groups of order $<2^{11}$ with center of order $2$, in possible ways to get group of order $2^{11}$.) $\endgroup$ – p Groups Mar 29 '16 at 8:53
  • 1
    $\begingroup$ At the end, I want center of order $2$. It seems, constructing random groups until it appears takes a very long time in average, so I probably need a method to rule out large centers to get such a group quickly. $\endgroup$ – Peter Mar 31 '16 at 22:20
0
$\begingroup$

This might be a way to achieve the desired goal, but I do not claim that it is the best way. (Whatever "best" might mean)

First, some work. Higman counted the number of $p$-class 2 groups (length of the lower exponent-$p$ series is 2) of order $p^n$ and concluded that there are at least $p^{2(n^3-6n^2)/27}$ such groups. Therefore, there are at least $2^{64}$ groups of order $4096$.

If you want a $2$-group with center of order $2$, then the $p$-class of the group can range from 3 to 12. If the $p$-class is 2, then the Frattini subgroup is a central subgroup, and in this case is equal to the center. Therefore, $G$ is $11$-generated and $G'=Z(G)=\Phi(G)$, so $G$ is extraspecial which cannot be true. Alternatively, the commutator map is a symplectic form of an odd dimensional space, and therefore, has a radical. Hence, the center has more than 2 elements.

If the $p$-class is $c\in\{3,...,12\}$, then $G$ is a central quotient of the $p$-covering group of $G/P_c(G)$. Specifically, if $P$ is the $p$-covering group of $G/P_c(G)$, then for some $H<Z(P)$, with $[Z(P):H]=2$, $G\cong P/H$. (Details: https://en.wikipedia.org/wiki/P-group_generation_algorithm)

The next issue one might consider is distribution. It seems that the larger the $p$-class, the fewer the number of groups bounded by a specific order. I don't think a statement like this has been proven, but there is data to back this is belief. Therefore, it seems most groups with a center of order 2 would be $p$-class 3. Hence, applying a uniform distribution to $p$-class would not result in a uniform distribution among the groups of order $4096$ and center of order $2$. I'm not sure if you care about this, but it's worth a mention.

With these ideas, one way to generate a random group would be the following (this is basically the $p$-group generation algorithm).

  1. Randomly choose the $p$-class: $c$.
  2. Randomly choose $c-1$ positive integers whose sum is $11$: $d_1,...,d_{c-1}$.
  3. Construct a random group $G$ of $p$-class $2$, order $d_1+d_2$, and generated by $d_1$ elements.
  4. Construct the $p$-covering group of $G$: $P$.
  5. Construct a random central subgroup $H$ of $P$, where $[Z(P):H]=d_3$, and set $G=P/H$.
  6. For all other $d_4,...,d_{c-1}$, do steps 4 and 5 again.
  7. Construct the $p$-covering group of $G$.
  8. Construct a random central subgroup $H$ of $P$, where $[Z(P):H]=1$.
  9. $G=P/H$.

It is possible that at some point in this, the group is terminal. That is, it is equal to its $p$-cover. If this happens, start over or backtrack.

To construct a random group of $p$-class 2, it is enough to construct $d_2$ random $d_1\times d_1$ matrices along with a quadratic map from a $d_1$ dimensional vector space into a $d_2$ dimensional vector space. I'm not sure if GAP has commands to do this, but if $d_1+d_2\leq 9$, then you can choose a random $p$-class 2 group from the databases in GAP and go from there.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.